Please see: https://groups.google.com/d/topic/bob-devel/l8XZHWhNNwg/discussion

I know that we have had the discussion, where to put the ZT-normalization, already long time ago, and that I was voting for having it inside bob.learn.em. However, I don't remember, why I voted for that, and I no longer think that this was a good decision.

In fact, I would rather assume to have the ZT-norm functionality inside bob.learn.linear -- as it contains linear transforms of scores -- though it does not use the linear machine. 
On the other hand, ZT-norm does not have anything to do with the EM algorithm, so it is completely lost in the bob.learn.em package.

So, what do you think about -- finally -- cleaning up bob.learn.em, which is the successor of the old bob.learn.misc package and finally move the ZT score normalization into the appropriate package?